Re: about IBS support on amd

On Thu, May 31, 2012 at 5:18 AM, Da Feng <jvfengda <at> yahoo.com> wrote:
> Hi:
> I want to sample IBS_OP_NB_REMOTE_DRAM events, and I find libpfm4 has ibs
> support in file pfmlib_amd64.c. But I can't find the ibs examples. Are there
> any or does libpfm support IBS_OP events?
>
The IBS support currently in libpfm4 if not operational. It does not
work with existing perf_events.
Current perf_event does not have IBS support. This is still work in progress.

> DaFENG
> Coder
> Telecommunication && Network Industry
> Gmail:sunspiderX <at> gmail.com
>
> ------------------------------------------------------------------------------
> Live Security Virtual Conference
> Exclusive live event will cover all the ways today's security and
> threat landscape has changed and how IT managers can respond. Discussions
> will include endpoint security, mobile security and the latest in malware
> threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
> _______________________________________________
> perfmon2-libpfm4-commits mailing list
> perfmon2-libpfm4-commits <at> lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/perfmon2-libpfm4-commits
>

------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and 

[PATCH] Should use siginfo_t rather than struct siginfo

Newer versions of glibc have removed struct siginfo. The
notify_group.c and notify_self.c code still used stuct siginfo.
This patch uses the siginfo_t instead.

Signed-off-by: William Cohen <wcohen <at> redhat.com>
---
 perf_examples/notify_group.c |    2 +-
 perf_examples/notify_self.c  |    2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/perf_examples/notify_group.c b/perf_examples/notify_group.c
index c6136c0..e3eb8d2 100644
--- a/perf_examples/notify_group.c
+++ b/perf_examples/notify_group.c
 <at>  <at>  -51,7 +51,7  <at>  <at>  static int num_fds;
 static int buffer_pages = 1; /* size of buffer payload  (must be power of 2) */

 static void
-sigio_handler(int n, struct siginfo *info, struct sigcontext *sc)
+sigio_handler(int n, siginfo_t *info, struct sigcontext *sc)
 {
 	struct perf_event_header ehdr;
 	uint64_t ip;
diff --git a/perf_examples/notify_self.c b/perf_examples/notify_self.c
index 85bb25a..da42aa0 100644
--- a/perf_examples/notify_self.c
+++ b/perf_examples/notify_self.c
 <at>  <at>  -47,7 +47,7  <at>  <at>  static int num_fds = 0;
 static int buffer_pages = 1; /* size of buffer payload (must be power of 2)*/

Re: [PATCH] Should use siginfo_t rather than struct siginfo

Papi folks, 

We should probably be testing this in autoconf. Lots of old releases have partial siginfo_ts so we need to
see which one to use. It also varies a bit by architecture.

Apologies for brevity and errors as this was sent from my mobile device.

On Jun 7, 2012, at 2:06 PM, William Cohen <wcohen <at> redhat.com> wrote:

> Newer versions of glibc have removed struct siginfo. The
> notify_group.c and notify_self.c code still used stuct siginfo.
> This patch uses the siginfo_t instead.
> 
> Signed-off-by: William Cohen <wcohen <at> redhat.com>
> ---
> perf_examples/notify_group.c |    2 +-
> perf_examples/notify_self.c  |    2 +-
> 2 files changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/perf_examples/notify_group.c b/perf_examples/notify_group.c
> index c6136c0..e3eb8d2 100644
> --- a/perf_examples/notify_group.c
> +++ b/perf_examples/notify_group.c
>  <at>  <at>  -51,7 +51,7  <at>  <at>  static int num_fds;
> static int buffer_pages = 1; /* size of buffer payload  (must be power of 2) */
> 
> static void
> -sigio_handler(int n, struct siginfo *info, struct sigcontext *sc)
> +sigio_handler(int n, siginfo_t *info, struct sigcontext *sc)
> {

libpfm4 IvyBridge support

Hi,

I have pushed the commit to add Ivy Bridge (model 58) support
to libpfm4.

Note that the kernel perf_event support is still missing. It will
appear in 3.5.0.

The event table is based on SDM Vol3 May 2012.

Thanks.

------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and 
threat landscape has changed and how IT managers can respond. Discussions 
will include endpoint security, mobile security and the latest in malware 
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/

Re: [PATCH] Should use siginfo_t rather than struct siginfo

Patch applied.
Thanks.

On Thu, Jun 7, 2012 at 9:06 PM, William Cohen <wcohen <at> redhat.com> wrote:
> Newer versions of glibc have removed struct siginfo. The
> notify_group.c and notify_self.c code still used stuct siginfo.
> This patch uses the siginfo_t instead.
>
> Signed-off-by: William Cohen <wcohen <at> redhat.com>
> ---
>  perf_examples/notify_group.c |    2 +-
>  perf_examples/notify_self.c  |    2 +-
>  2 files changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/perf_examples/notify_group.c b/perf_examples/notify_group.c
> index c6136c0..e3eb8d2 100644
> --- a/perf_examples/notify_group.c
> +++ b/perf_examples/notify_group.c
>  <at>  <at>  -51,7 +51,7  <at>  <at>  static int num_fds;
>  static int buffer_pages = 1; /* size of buffer payload  (must be power of 2) */
>
>  static void
> -sigio_handler(int n, struct siginfo *info, struct sigcontext *sc)
> +sigio_handler(int n, siginfo_t *info, struct sigcontext *sc)
>  {
>        struct perf_event_header ehdr;
>        uint64_t ip;
> diff --git a/perf_examples/notify_self.c b/perf_examples/notify_self.c
> index 85bb25a..da42aa0 100644
> --- a/perf_examples/notify_self.c

swig error for libpfm build

I am in the process of getting libpfm into fedora rawhide.  When building with the older version of swig in
rawhide (2.0.6) there was the following message in the build:

swig -python -I../include -o src/perfmon_int_wrap.c src/perfmon_int.i
src/perfmon_int.i:95: Warning 521: Illegal destructor name ~perf_event_attr_t. Ignored.

For the newer version of swig in rawhide (2.0.7) swig gets a segmentation fault. I have filed the following
bug to get swig fixed:

https://bugzilla.redhat.com/show_bug.cgi?id=830249

In the meantime it would be good to get libpfm upstream fixed so swig doesn't complain about this error. I
think that the attached patch eliminates this problem.  This is based on the approach mentioned in a
similar swig bug report:

http://sourceforge.net/tracker/?func=detail&aid=3530118&group_id=1645&atid=101645

-Will

------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and 
threat landscape has changed and how IT managers can respond. Discussions 
will include endpoint security, mobile security and the latest in malware 
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/

_______________________________________________
perfmon2-devel mailing list

Re: swig error for libpfm build

On 6/8/12 9:46 AM, William Cohen wrote:
> In the meantime it would be good to get libpfm upstream fixed so swig doesn't complain about this error. I
think that the attached patch eliminates this problem.  This is based on the approach mentioned in a
similar swig bug report:
>
> http://sourceforge.net/tracker/?func=detail&aid=3530118&group_id=1645&atid=101645
>

Yes, this is getting covered via:

%typemap(argout) T* output

Looks good to me.

  -Arun

------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and 
threat landscape has changed and how IT managers can respond. Discussions 
will include endpoint security, mobile security and the latest in malware 
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/

Re: [PATCH] Should use siginfo_t rather than struct siginfo

On 06/07/2012 03:42 PM, Philip Mucci wrote:
> Papi folks, 
> 
> We should probably be testing this in autoconf. Lots of old releases have partial siginfo_ts so we need to
see which one to use. It also varies a bit by architecture.
> 
> Apologies for brevity and errors as this was sent from my mobile device.

Hi Phil,

I notice that papi encounters a similar problem.  What would be requirement for the autoconf check? Use the
struct siginfo if it exists and siginfo_t if no struct siginfo?

-Will
> 
> On Jun 7, 2012, at 2:06 PM, William Cohen <wcohen <at> redhat.com> wrote:
> 
>> Newer versions of glibc have removed struct siginfo. The
>> notify_group.c and notify_self.c code still used stuct siginfo.
>> This patch uses the siginfo_t instead.
>>
>> Signed-off-by: William Cohen <wcohen <at> redhat.com>
>> ---
>> perf_examples/notify_group.c |    2 +-
>> perf_examples/notify_self.c  |    2 +-
>> 2 files changed, 2 insertions(+), 2 deletions(-)
>>
>> diff --git a/perf_examples/notify_group.c b/perf_examples/notify_group.c
>> index c6136c0..e3eb8d2 100644
>> --- a/perf_examples/notify_group.c

Re: [PATCH] Should use siginfo_t rather than struct siginfo

On 06/11/2012 04:15 PM, Vince Weaver wrote:
> On Mon, 11 Jun 2012, William Cohen wrote:
> 
>> Hi Phil,
>>
>> I notice that papi encounters a similar problem.  What would be requirement for the autoconf check? Use
the struct siginfo if it exists and siginfo_t if no struct siginfo?
>>
>> -Will
> 
> 
> From what I gather the siginfo_t / struct siginfo stuff was only broken in 
> Linux 2.4.  I don't think PAPI really supports anything that old anymore.
> 
> It looks like all of our architectures besides Linux (at least the ones 
> that don't just have a dummy hwd_siginfo_t) are using siginfo_t already.
> 
> The change is just a one line fix in linux-context.h.  I'm willing to 
> commit it and back it out if it breaks anything.
> 
> Vince
> 

Hi Vince,

I have a one line fix that allow papi to compile with the siginfo_t. The patch is attached.

-Will

Re: [PATCH] Should use siginfo_t rather than struct siginfo

On Mon, 11 Jun 2012, William Cohen wrote:

> Hi Phil,
> 
> I notice that papi encounters a similar problem.  What would be requirement for the autoconf check? Use the
struct siginfo if it exists and siginfo_t if no struct siginfo?
> 
> -Will

>From what I gather the siginfo_t / struct siginfo stuff was only broken in 
Linux 2.4.  I don't think PAPI really supports anything that old anymore.

It looks like all of our architectures besides Linux (at least the ones 
that don't just have a dummy hwd_siginfo_t) are using siginfo_t already.

The change is just a one line fix in linux-context.h.  I'm willing to 
commit it and back it out if it breaks anything.

Vince

------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and 
threat landscape has changed and how IT managers can respond. Discussions 
will include endpoint security, mobile security and the latest in malware 
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/